i965: Delete intel_context entirely.
[mesa.git] / src / mesa / drivers / dri / i965 / gen6_sf_state.c
1 /*
2 * Copyright © 2009 Intel Corporation
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
10 *
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
13 * Software.
14 *
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
21 * IN THE SOFTWARE.
22 *
23 * Authors:
24 * Eric Anholt <eric@anholt.net>
25 *
26 */
27
28 #include "brw_context.h"
29 #include "brw_state.h"
30 #include "brw_defines.h"
31 #include "brw_util.h"
32 #include "main/macros.h"
33 #include "main/fbobject.h"
34 #include "intel_batchbuffer.h"
35
36 /**
37 * Determine the appropriate attribute override value to store into the
38 * 3DSTATE_SF structure for a given fragment shader attribute. The attribute
39 * override value contains two pieces of information: the location of the
40 * attribute in the VUE (relative to urb_entry_read_offset, see below), and a
41 * flag indicating whether to "swizzle" the attribute based on the direction
42 * the triangle is facing.
43 *
44 * If an attribute is "swizzled", then the given VUE location is used for
45 * front-facing triangles, and the VUE location that immediately follows is
46 * used for back-facing triangles. We use this to implement the mapping from
47 * gl_FrontColor/gl_BackColor to gl_Color.
48 *
49 * urb_entry_read_offset is the offset into the VUE at which the SF unit is
50 * being instructed to begin reading attribute data. It can be set to a
51 * nonzero value to prevent the SF unit from wasting time reading elements of
52 * the VUE that are not needed by the fragment shader. It is measured in
53 * 256-bit increments.
54 */
55 uint32_t
56 get_attr_override(const struct brw_vue_map *vue_map, int urb_entry_read_offset,
57 int fs_attr, bool two_side_color, uint32_t *max_source_attr)
58 {
59 if (fs_attr == VARYING_SLOT_POS) {
60 /* This attribute will be overwritten by the fragment shader's
61 * interpolation code (see emit_interp() in brw_wm_fp.c), so just let it
62 * reference the first available attribute.
63 */
64 return 0;
65 }
66
67 /* Find the VUE slot for this attribute. */
68 int slot = vue_map->varying_to_slot[fs_attr];
69
70 /* If there was only a back color written but not front, use back
71 * as the color instead of undefined
72 */
73 if (slot == -1 && fs_attr == VARYING_SLOT_COL0)
74 slot = vue_map->varying_to_slot[VARYING_SLOT_BFC0];
75 if (slot == -1 && fs_attr == VARYING_SLOT_COL1)
76 slot = vue_map->varying_to_slot[VARYING_SLOT_BFC1];
77
78 if (slot == -1) {
79 /* This attribute does not exist in the VUE--that means that the vertex
80 * shader did not write to it. This means that either:
81 *
82 * (a) This attribute is a texture coordinate, and it is going to be
83 * replaced with point coordinates (as a consequence of a call to
84 * glTexEnvi(GL_POINT_SPRITE, GL_COORD_REPLACE, GL_TRUE)), so the
85 * hardware will ignore whatever attribute override we supply.
86 *
87 * (b) This attribute is read by the fragment shader but not written by
88 * the vertex shader, so its value is undefined. Therefore the
89 * attribute override we supply doesn't matter.
90 *
91 * In either case the attribute override we supply doesn't matter, so
92 * just reference the first available attribute.
93 */
94 return 0;
95 }
96
97 /* Compute the location of the attribute relative to urb_entry_read_offset.
98 * Each increment of urb_entry_read_offset represents a 256-bit value, so
99 * it counts for two 128-bit VUE slots.
100 */
101 int source_attr = slot - 2 * urb_entry_read_offset;
102 assert(source_attr >= 0 && source_attr < 32);
103
104 /* If we are doing two-sided color, and the VUE slot following this one
105 * represents a back-facing color, then we need to instruct the SF unit to
106 * do back-facing swizzling.
107 */
108 bool swizzling = two_side_color &&
109 ((vue_map->slot_to_varying[slot] == VARYING_SLOT_COL0 &&
110 vue_map->slot_to_varying[slot+1] == VARYING_SLOT_BFC0) ||
111 (vue_map->slot_to_varying[slot] == VARYING_SLOT_COL1 &&
112 vue_map->slot_to_varying[slot+1] == VARYING_SLOT_BFC1));
113
114 /* Update max_source_attr. If swizzling, the SF will read this slot + 1. */
115 if (*max_source_attr < source_attr + swizzling)
116 *max_source_attr = source_attr + swizzling;
117
118 if (swizzling) {
119 return source_attr |
120 (ATTRIBUTE_SWIZZLE_INPUTATTR_FACING << ATTRIBUTE_SWIZZLE_SHIFT);
121 }
122
123 return source_attr;
124 }
125
126 static void
127 upload_sf_state(struct brw_context *brw)
128 {
129 struct gl_context *ctx = &brw->ctx;
130 /* BRW_NEW_FRAGMENT_PROGRAM */
131 uint32_t num_outputs = _mesa_bitcount_64(brw->fragment_program->Base.InputsRead);
132 /* _NEW_LIGHT */
133 bool shade_model_flat = ctx->Light.ShadeModel == GL_FLAT;
134 uint32_t dw1, dw2, dw3, dw4, dw16, dw17;
135 int i;
136 /* _NEW_BUFFER */
137 bool render_to_fbo = _mesa_is_user_fbo(ctx->DrawBuffer);
138 bool multisampled_fbo = ctx->DrawBuffer->Visual.samples > 1;
139
140 int attr = 0, input_index = 0;
141 int urb_entry_read_offset = 1;
142 float point_size;
143 uint16_t attr_overrides[VARYING_SLOT_MAX];
144 uint32_t point_sprite_origin;
145
146 dw1 = GEN6_SF_SWIZZLE_ENABLE | num_outputs << GEN6_SF_NUM_OUTPUTS_SHIFT;
147
148 dw2 = GEN6_SF_STATISTICS_ENABLE |
149 GEN6_SF_VIEWPORT_TRANSFORM_ENABLE;
150
151 dw3 = 0;
152 dw4 = 0;
153 dw16 = 0;
154 dw17 = 0;
155
156 /* _NEW_POLYGON */
157 if ((ctx->Polygon.FrontFace == GL_CCW) ^ render_to_fbo)
158 dw2 |= GEN6_SF_WINDING_CCW;
159
160 if (ctx->Polygon.OffsetFill)
161 dw2 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_SOLID;
162
163 if (ctx->Polygon.OffsetLine)
164 dw2 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_WIREFRAME;
165
166 if (ctx->Polygon.OffsetPoint)
167 dw2 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_POINT;
168
169 switch (ctx->Polygon.FrontMode) {
170 case GL_FILL:
171 dw2 |= GEN6_SF_FRONT_SOLID;
172 break;
173
174 case GL_LINE:
175 dw2 |= GEN6_SF_FRONT_WIREFRAME;
176 break;
177
178 case GL_POINT:
179 dw2 |= GEN6_SF_FRONT_POINT;
180 break;
181
182 default:
183 assert(0);
184 break;
185 }
186
187 switch (ctx->Polygon.BackMode) {
188 case GL_FILL:
189 dw2 |= GEN6_SF_BACK_SOLID;
190 break;
191
192 case GL_LINE:
193 dw2 |= GEN6_SF_BACK_WIREFRAME;
194 break;
195
196 case GL_POINT:
197 dw2 |= GEN6_SF_BACK_POINT;
198 break;
199
200 default:
201 assert(0);
202 break;
203 }
204
205 /* _NEW_SCISSOR */
206 if (ctx->Scissor.Enabled)
207 dw3 |= GEN6_SF_SCISSOR_ENABLE;
208
209 /* _NEW_POLYGON */
210 if (ctx->Polygon.CullFlag) {
211 switch (ctx->Polygon.CullFaceMode) {
212 case GL_FRONT:
213 dw3 |= GEN6_SF_CULL_FRONT;
214 break;
215 case GL_BACK:
216 dw3 |= GEN6_SF_CULL_BACK;
217 break;
218 case GL_FRONT_AND_BACK:
219 dw3 |= GEN6_SF_CULL_BOTH;
220 break;
221 default:
222 assert(0);
223 break;
224 }
225 } else {
226 dw3 |= GEN6_SF_CULL_NONE;
227 }
228
229 /* _NEW_LINE */
230 {
231 uint32_t line_width_u3_7 = U_FIXED(CLAMP(ctx->Line.Width, 0.0, 7.99), 7);
232 /* TODO: line width of 0 is not allowed when MSAA enabled */
233 if (line_width_u3_7 == 0)
234 line_width_u3_7 = 1;
235 dw3 |= line_width_u3_7 << GEN6_SF_LINE_WIDTH_SHIFT;
236 }
237 if (ctx->Line.SmoothFlag) {
238 dw3 |= GEN6_SF_LINE_AA_ENABLE;
239 dw3 |= GEN6_SF_LINE_AA_MODE_TRUE;
240 dw3 |= GEN6_SF_LINE_END_CAP_WIDTH_1_0;
241 }
242 /* _NEW_MULTISAMPLE */
243 if (multisampled_fbo && ctx->Multisample.Enabled)
244 dw3 |= GEN6_SF_MSRAST_ON_PATTERN;
245
246 /* _NEW_PROGRAM | _NEW_POINT */
247 if (!(ctx->VertexProgram.PointSizeEnabled ||
248 ctx->Point._Attenuated))
249 dw4 |= GEN6_SF_USE_STATE_POINT_WIDTH;
250
251 /* Clamp to ARB_point_parameters user limits */
252 point_size = CLAMP(ctx->Point.Size, ctx->Point.MinSize, ctx->Point.MaxSize);
253
254 /* Clamp to the hardware limits and convert to fixed point */
255 dw4 |= U_FIXED(CLAMP(point_size, 0.125, 255.875), 3);
256
257 /*
258 * Window coordinates in an FBO are inverted, which means point
259 * sprite origin must be inverted, too.
260 */
261 if ((ctx->Point.SpriteOrigin == GL_LOWER_LEFT) != render_to_fbo) {
262 point_sprite_origin = GEN6_SF_POINT_SPRITE_LOWERLEFT;
263 } else {
264 point_sprite_origin = GEN6_SF_POINT_SPRITE_UPPERLEFT;
265 }
266 dw1 |= point_sprite_origin;
267
268 /* _NEW_LIGHT */
269 if (ctx->Light.ProvokingVertex != GL_FIRST_VERTEX_CONVENTION) {
270 dw4 |=
271 (2 << GEN6_SF_TRI_PROVOKE_SHIFT) |
272 (2 << GEN6_SF_TRIFAN_PROVOKE_SHIFT) |
273 (1 << GEN6_SF_LINE_PROVOKE_SHIFT);
274 } else {
275 dw4 |=
276 (1 << GEN6_SF_TRIFAN_PROVOKE_SHIFT);
277 }
278
279 /* Create the mapping from the FS inputs we produce to the VS outputs
280 * they source from.
281 */
282 uint32_t max_source_attr = 0;
283 for (; attr < VARYING_SLOT_MAX; attr++) {
284 enum glsl_interp_qualifier interp_qualifier =
285 brw->fragment_program->InterpQualifier[attr];
286 bool is_gl_Color = attr == VARYING_SLOT_COL0 || attr == VARYING_SLOT_COL1;
287
288 if (!(brw->fragment_program->Base.InputsRead & BITFIELD64_BIT(attr)))
289 continue;
290
291 /* _NEW_POINT */
292 if (ctx->Point.PointSprite &&
293 (attr >= VARYING_SLOT_TEX0 && attr <= VARYING_SLOT_TEX7) &&
294 ctx->Point.CoordReplace[attr - VARYING_SLOT_TEX0]) {
295 dw16 |= (1 << input_index);
296 }
297
298 if (attr == VARYING_SLOT_PNTC)
299 dw16 |= (1 << input_index);
300
301 /* flat shading */
302 if (interp_qualifier == INTERP_QUALIFIER_FLAT ||
303 (shade_model_flat && is_gl_Color &&
304 interp_qualifier == INTERP_QUALIFIER_NONE))
305 dw17 |= (1 << input_index);
306
307 /* The hardware can only do the overrides on 16 overrides at a
308 * time, and the other up to 16 have to be lined up so that the
309 * input index = the output index. We'll need to do some
310 * tweaking to make sure that's the case.
311 */
312 assert(input_index < 16 || attr == input_index);
313
314 /* BRW_NEW_VUE_MAP_GEOM_OUT | _NEW_LIGHT | _NEW_PROGRAM */
315 attr_overrides[input_index++] =
316 get_attr_override(&brw->vue_map_geom_out,
317 urb_entry_read_offset, attr,
318 ctx->VertexProgram._TwoSideEnabled,
319 &max_source_attr);
320 }
321
322 for (; input_index < VARYING_SLOT_MAX; input_index++)
323 attr_overrides[input_index] = 0;
324
325 /* From the Sandy Bridge PRM, Volume 2, Part 1, documentation for
326 * 3DSTATE_SF DWord 1 bits 15:11, "Vertex URB Entry Read Length":
327 *
328 * "This field should be set to the minimum length required to read the
329 * maximum source attribute. The maximum source attribute is indicated
330 * by the maximum value of the enabled Attribute # Source Attribute if
331 * Attribute Swizzle Enable is set, Number of Output Attributes-1 if
332 * enable is not set.
333 * read_length = ceiling((max_source_attr + 1) / 2)
334 *
335 * [errata] Corruption/Hang possible if length programmed larger than
336 * recommended"
337 */
338 uint32_t urb_entry_read_length = ALIGN(max_source_attr + 1, 2) / 2;
339 dw1 |= urb_entry_read_length << GEN6_SF_URB_ENTRY_READ_LENGTH_SHIFT |
340 urb_entry_read_offset << GEN6_SF_URB_ENTRY_READ_OFFSET_SHIFT;
341
342 BEGIN_BATCH(20);
343 OUT_BATCH(_3DSTATE_SF << 16 | (20 - 2));
344 OUT_BATCH(dw1);
345 OUT_BATCH(dw2);
346 OUT_BATCH(dw3);
347 OUT_BATCH(dw4);
348 OUT_BATCH_F(ctx->Polygon.OffsetUnits * 2); /* constant. copied from gen4 */
349 OUT_BATCH_F(ctx->Polygon.OffsetFactor); /* scale */
350 OUT_BATCH_F(0.0); /* XXX: global depth offset clamp */
351 for (i = 0; i < 8; i++) {
352 OUT_BATCH(attr_overrides[i * 2] | attr_overrides[i * 2 + 1] << 16);
353 }
354 OUT_BATCH(dw16); /* point sprite texcoord bitmask */
355 OUT_BATCH(dw17); /* constant interp bitmask */
356 OUT_BATCH(0); /* wrapshortest enables 0-7 */
357 OUT_BATCH(0); /* wrapshortest enables 8-15 */
358 ADVANCE_BATCH();
359 }
360
361 const struct brw_tracked_state gen6_sf_state = {
362 .dirty = {
363 .mesa = (_NEW_LIGHT |
364 _NEW_PROGRAM |
365 _NEW_POLYGON |
366 _NEW_LINE |
367 _NEW_SCISSOR |
368 _NEW_BUFFERS |
369 _NEW_POINT |
370 _NEW_MULTISAMPLE),
371 .brw = (BRW_NEW_CONTEXT |
372 BRW_NEW_FRAGMENT_PROGRAM |
373 BRW_NEW_VUE_MAP_GEOM_OUT)
374 },
375 .emit = upload_sf_state,
376 };